ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2019, Vol. 56 ›› Issue (3): 635-642.doi: 10.7544/issn1000-1239.2019.20170807

• 人工智能 • 上一篇    下一篇

多元数据融合的非干扰身份识别方法

于佃存1,2,陈益强1,3,彭晓晖1,焦帅1,李啸海1,钟习1,3   

  1. 1(中国科学院计算技术研究所 北京 100190); 2(山东大学软件学院 济南 250101); 3(中国科学院大学 北京 100049) (yudiancun@ict.ac.cn)
  • 出版日期: 2019-03-01
  • 基金资助: 
    国家重点研发计划基金项目(2017YFB1002801);国家自然科学基金项目(61572471);广东省科技计划项目(2015B010105001);中国科学院率先行动“百人计划”项目(Y704061000)

Multi-Model Data Fusion Based Unobtrusive Identification Method

Yu Diancun1,2, Chen Yiqiang1,3, Peng Xiaohui1, Jiao Shuai1, Li Xiaohai1, Zhong Xi1,3   

  1. 1(Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190); 2(School of Software, Shandong University, Jinan 250101); 3(University of Chinese Academy of Sciences, Beijing 100049)
  • Online: 2019-03-01

摘要: 传统的步态识别技术在智能终端设备风险控制领域的应用还存在一些问题,已有的方案是通过加速度、陀螺仪等多传感器对步态进行身份识别与验证.由于现有的识别方法设置了许多限制条件,给该技术的使用与推广俱造成了困难.例如:需要把传感器设备固定在脚踝、膝盖、腰部等位置,设备有指定的朝向,用户做特定的动作.通过步态进行身份识别与验证的技术应用到风险控制领域需要一套完整可靠的系统架构,现有架构还存在较大问题.因此,提出一种与位置、行为无关的非干扰的身份识别与验证方法,该方法仅使用加速度传感器,并以此方法为核心建立了一套完整的系统实现架构,该架构方法的实现提高了系统的整体精度与可用性.首先对用户的行为及设备所在的位置进行预测;然后针对性地进行步态分析与识别.实验中仅使用智能手机中内置的加速度传感器采集数据,最后对步态进行位置无关的分析与识别最重确定用户身份,从而起到降低智能手机使用风险提高安全系数的作用.实验结果表明设计的系统架构有利于系统整体精度的提升,且该方法具有较高的识别率和极低的假阳率(false positive rate, FPR),且在非干扰用户的情况下提高了APP和智能手机等智能终端设备的安全性.

关键词: 人机交互, 安全, 步态, 风险控制, 非干扰

Abstract: The traditional gait recognition technology in the field of intelligent terminal equipment risk control still has some problems. The existing program is using accelerometer, gyroscope and other multi-sensor gait for identification and verification. Due to the existing identification methods set a number of restrictions, hinder the use and promotion of this technology. For example: the sensor device needs to be fixed at the same position as the ankle, knee, waist and so on; the device has a designated orientation; the user does a specific action. In addition, the application of the technology of identity verification and verification through gait to the field of risk control requires a complete and reliable system architecture. There is still a big problem with the existing architecture. Therefore, this paper presents a non-interference and location-independent identification and verification method that uses only accelerometers and builds a complete set of system implementation architecture with this method as the core. The implementation of this architecture method has improved the overall system accuracy and availability. Firstly, the user’s behavior and the location of the device are predicted; then the gait analysis and identification are carried out. In this experiment, we only use the built-in accelerometer in the smart phone to collect data, finally position-independent gait analysis and identification to identify the user to determine which is the most important, so as to reduce the risk of using smart phones and improve the safety factor. The experimental results show that the system architecture designed in this paper is conducive to the improvement of overall system accuracy. The method has the characteristics of high recognition rate and very low FPR (false positive rate), and improves the APP and the smartphone in the case of non-interfering users such as intelligent terminal equipment security.

Key words: human-computer interaction, safety, gait, risk control, noninterference

中图分类号: